// Analyze OpenClaw, Claude Cowork, Hermes Agent, and adjacent agent products to learn what to borrow, what to directly copy as patterns, and what Spark Intelligence should keep uniquely its own. Use when comparing architectures, onboarding, adapters, reliability harnesses, migration paths, or competitive product shape.
Audit local-agent systems, specs, and code for security weaknesses and hardening priorities. Use when reviewing Spark Intelligence features, security-sensitive diffs, identity or pairing flows, host-execution boundaries, webhook adapters, auth and secret handling, or when producing a security grade and remediation list that others can reuse later.
Design secure, lightweight local-agent systems for Spark Intelligence and adjacent agent products. Use when defining identity and pairing rules, provider or channel auth, dangerous command approvals, tool and host boundaries, webhook trust checks, local secret handling, or security guardrails for new features.
Build and review cron jobs, schedulers, retries, smoke tests, and operational harnesses for Spark Intelligence in a lightweight and highly reliable way. Use when designing job systems, background work, health checks, retry logic, cron behavior, migration jobs, or smoke-test coverage for scheduled work.
Make Spark Intelligence product decisions with deep awareness of Spark Researcher, Spark Swarm, domain chips, specialization paths, autoloop flywheels, and the wider Spark ecosystem. Use when defining product scope, modular architecture, ecosystem integration, persistent agent behavior, or self-evolving collective-intelligence systems.
Keep Spark Intelligence code, architecture, security, and documentation highly maintainable, scalable enough, and clean without adding unnecessary complexity. Use when reviewing code, planning architecture, writing docs, simplifying modules, improving quality, or enforcing long-term engineering standards.
| name | agent-landscape-analysis |
| description | Analyze OpenClaw, Claude Cowork, Hermes Agent, and adjacent agent products to learn what to borrow, what to directly copy as patterns, and what Spark Intelligence should keep uniquely its own. Use when comparing architectures, onboarding, adapters, reliability harnesses, migration paths, or competitive product shape. |
Use this skill when work touches:
docs/ARCHITECTURE_SPARK_INTELLIGENCE_V1.mddocs/CRON_JOB_HARNESS_SPEC_V1.mddocs/IMPORT_AND_MIGRATION_SPEC_V1.mddocs/CODING_RULESET_V1.mddocs/SPARK_INTELLIGENCE_PROMPT_BIBLE.mdThen read:
references/comparison-playbook.mdThe goal is not imitation.
The goal is disciplined comparison that helps Spark Intelligence become:
Prefer:
Use primary sources whenever possible:
Treat Claude Cowork as a product and UX benchmark unless trustworthy technical internals are publicly available. Do not invent proprietary implementation details.
Always classify findings into:
Use the comparison buckets and output template in references/comparison-playbook.md.
Return these explicitly:
The result should end with: